Factories Plan Bigger Budgets for 2027. Half Cannot Show What Their AI Returned

Capital planning season has started, and manufacturers are going into it confident. Regional factory surveys published this week show activity still expanding, and the firms questioned say they intend to spend more next year, not less. Software and computing are a large part of that plan. The awkward part is what comes before the new money is approved: a credible account of what the last round of digital spending, and in particular the rush into AI, actually delivered. On the latest evidence, about half of manufacturers cannot yet give one.

Nobody in the Panel Is Planning to Cut

The Philadelphia Fed's October Manufacturing Business Outlook Survey, released on October 15, asked firms how their 2027 capital expenditures would compare with 2026. For total capital spending, 52.2% said higher, 47.8% said the same and none said lower. In the bank's own words, "no firm expects lower capital spending next year."

The detail is just as telling. Expectations of higher spending were most widespread for noncomputer equipment, where half of firms expect to spend more. Software and computer hardware followed, each with 39.1% expecting increases and 8.7% expecting cuts. Structures came next at 34.8%. Energy-saving investments ranked lowest of the named categories, with 27.3% planning to spend more and 13.6% planning to spend less.

That confidence is arriving while near-term activity cools in the region. The survey's general activity index fell from 37.8 in September to 20.9 in October, and the new orders index dropped from 29.2 to 4.3. The index for future capital expenditures edged down 2 points to 34.9, still well into positive territory. Further north the picture was stronger. The New York Fed's Empire State Manufacturing Survey reported its general business conditions index up 18 points to 25.5, the highest level in nearly five years, with new orders up 21 points to 23.2. Prices paid remain high in both surveys, at 43.2 in Philadelphia and 59.5 in New York.

Put together, the message for finance teams is that budgets for 2027 will grow, but input costs will keep pressing on margin while they do. That makes every capital line harder to defend on optimism alone, and software lines most of all, because their payback is the least visible.

AI Adoption Has Outrun the Measurement

The clearest recent read on that problem comes from research released by Revalize on September 29. The software company surveyed 500 decision makers at manufacturers with at least 100 employees, 225 in the US, 225 in Germany, Austria and Switzerland and 50 in the UK, between July 7 and July 17. Respondents were people responsible for configure-price-quote, product lifecycle and engineering simulation software, so the sample leans towards engineering and commercial systems rather than the plant floor.

Adoption has moved quickly. Revalize reports that 36% of manufacturers had reached broad or advanced AI adoption in the first quarter of 2026, and that six months later 60% had AI integrated across multiple operations or fully embedded in their work. Only 8% remain in initial exploration.

Measurement has not kept pace. Only 49% of manufacturers measure AI return on investment rigorously, and 50% track AI outcomes inconsistently or informally. At the same time 72% say they can attribute business outcomes to AI, which suggests many firms believe the technology is working without the numbers to show a finance committee. The survey also found C-suite leaders more than twice as likely as senior managers to report advanced AI adoption, a reminder that the view from the top of the organisation is usually the more generous one.

There is a cost argument driving the spending as well. Revalize found that 44% of manufacturers are accelerating AI adoption to offset tariff and cost pressure. If AI is being bought as a cost defence, its returns should be measurable in exactly the terms a cost defence is judged by: hours, scrap, quote turnaround and margin. Where those numbers are missing, the investment is resting on belief.

Why the Gap Matters for the 2027 Budget

A capital committee faced with a 2027 request for more software has three questions. What did last year's spend deliver, how do we know, and why will the next tranche do better? A plant that tracks AI outcomes informally can answer the first question with anecdotes, the second barely at all and the third only with a vendor's forecast.

That matters more when equipment is competing for the same money. Half of the Philadelphia Fed panel expects to spend more on noncomputer equipment, the most widespread increase in the survey. Machines come with nameplate capacity and a quoted payback; software projects often come with neither. When budgets tighten, the line with the weaker evidence is the one that gets cut, regardless of which would have returned more.

The fix is not complicated, but it has to start before the money is spent. Each digital project needs a baseline, a named owner and a small set of operational measures agreed with finance in advance. Measured that way, a good project can defend its own renewal.

A Budget Season Playbook for Software and AI

• Inventory what is already running: list every AI and software initiative live today, its owner and the problem it was bought to solve before asking for anything new.
• Set baselines before go-live: record the current scrap rate, downtime, quote cycle or labour hours a project targets, or the improvement can never be proven.
• Agree the measures with finance: choose two or three operational metrics per project that the finance team accepts as evidence, and report them on the same cycle as the budget.
• Ask the floor, not just the boardroom: senior managers and supervisors see adoption differently from executives, so test claims of advanced use against the people who run the process.
• Fund scale-up on results: release 2027 money for expansion in stages, tied to measured outcomes from the first site or line.
• Do not let energy savings fall off the list: energy-saving investment ranked last in the Philadelphia Fed survey, yet it carries one of the most measurable paybacks a plant can offer.

The 2027 budgets look set to grow, and that is good news for plants that need new equipment and better systems. The firms that get the most from them will be the ones that can show, line by line, what this year's spending returned.